nvme-rdma: fix possible use-after-free in transport error_recovery work
authorSagi Grimberg <sagi@grimberg.me>
Tue, 1 Feb 2022 12:54:21 +0000 (14:54 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 23 Feb 2022 11:03:07 +0000 (12:03 +0100)
commit646952b2210f19e584d2bf9eb5d092abdca2fcc1
tree70890b2eeac302ce832f0b2cb881c28da19f33f4
parent5e42fca37ccc76f39f73732661bd47254cad5982
nvme-rdma: fix possible use-after-free in transport error_recovery work

[ Upstream commit b6bb1722f34bbdbabed27acdceaf585d300c5fd2 ]

While nvme_rdma_submit_async_event_work is checking the ctrl and queue
state before preparing the AER command and scheduling io_work, in order
to fully prevent a race where this check is not reliable the error
recovery work must flush async_event_work before continuing to destroy
the admin queue after setting the ctrl state to RESETTING such that
there is no race .submit_async_event and the error recovery handler
itself changing the ctrl state.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/nvme/host/rdma.c